Modelling Cosmic Radiation Events
in the Tree-Ring Radiocarbon Record
Benjamin Pope, UQ
Slides available at
UQ Students Jordan Dennis, Qingyuan Zhang, Utkarsh Sharma
Miyake Events
Tree rings give a "calibration curve" history of radiocarbon in the atmosphere
774-775 AD
Miyake et al, 2012
Anglo-Saxon Chronicle: 774 AD
“This year also appeared in the heavens a red crucifix, after sunset; the Mercians and the men of Kent fought at Otford; and wonderful serpents were seen in the land of the South-Saxons.”
Allen, 2012
Also in 993-994 AD
Miyake et al, 2013
Four more now known:
660 BC, 5259 BC, 5410 BC, 7176 BC
Vikings in North America!
Leif Eriksson, 'Leif the Lucky' from Iceland, landed in Vínland far to the West around 1000 AD.
Was this North America? Ruins at L'Anse aux Meadows in Newfoundland consistent with this!
Dated to 1021 CE by tree rings containing the 993 event!
So what is the physics?
Supernovae?
No!
Dee, Pope et al., 2016a
Magnetars?
Connection to fast radio bursts?
The Sun
Carrington Event
F.E. Church, "Aurora Borealis", 1865.
Superflares in Kepler
Maehara et al., 2012
Our Work
To model these time series, you have to model the entire global carbon cycle!
Carbon 'box models' with systems of transport equations between carbon reservoirs
(eg Güttler et al, 2015 model)
The problem is, none of the work in the field is reproducible! All these carbon box models are closed-source.
But we want to do astrophysics - infer event amplitudes, durations...
Built open-source Python package
Features:
- Literature carbon box models
- Built on Google Jax for autodiff + speed
- Multiple, fast (~10ms) numerical and analytic ODE solvers
- Nonparametric inverse solvers, using GPs or an exact ODE inverse
- MCMC and Nested Sampling support for Bayesian parameter inference
Example Posterior Draws
Example Posterior Distribution
Uusitalo et al, 2018 claim trend with latitude
With more data we find no trend
Amplitude of the events - equivalent years
Relation to the solar cycle
Long duration?
